Key Points
- AI is a novel technology that isn’t specifically contemplated by the U.K. tax rules.
- Careful tax due diligence is essential for both buyers and sellers of AI businesses in the U.K.
- Key tax breaks do not always clearly apply to AI businesses.
- An exemption from U.K. tax on share sales may not apply if the business is not considered a “trading” (as opposed to investment) company — potentially resulting in a significant tax bill for sellers.
- Autonomous AI agents raise interesting tax residency questions that have not yet been tested.
__________
When artificial intelligence (AI) businesses come to market, the valuation narrative typically is clear: a differentiated model, defensible data and scale within easy reach. But the tax analysis is often not clear. The U.K. tax rules were designed around patents, plants and people, not machine learning systems trained on diffuse, and sometimes legally ambiguous, inputs.
The resulting gap between what AI companies think they own, and what the tax legislation recognizes, may be significant — and surprising — for buyers of AI businesses that have ties to the U.K., and the law remains largely untested.
The Data Question
In many AI deals, the dataset is a key asset. It’s what the model was trained on and what makes the commercial output unique. However, U.K. tax law does not recognize data as a stand-alone asset class. Data assets need to be mapped onto existing legal categories — copyright, database rights, know-how or contractual rights.
Data assembled through substantial investment in collection, verification or presentation may qualify for sui generis database rights — a recognized intellectual property (IP) right that sits neatly within the U.K. tax regime. However, some AI training datasets — particularly those assembled through scraping or aggregation with limited human curation — will struggle to meet that threshold. Where no enforceable IP right exists, the value may instead be goodwill or other unregistered intangibles, and tax relief for those categories has been significantly restricted since 2015.
Licensed datasets raise a different problem: The value sits in the contract, not the content. The tax analysis follows the terms and accounting treatment of the contracts. Proprietary operational data (e.g., user behavior, telemetry transaction data) may constitute know-how or trade secrets, but only to the extent that control and protection can be evidenced.
This classification determines:
- Whether the buyer gets tax relief.
- How the asset is treated in a group reorganization.
- Whether any U.K. incentives for innovation such as research and development (R&D) tax credits or special low tax rate on patent income (the so-called Patent Box) are available.
The Exit Risk Can Be Substantial
For U.K. corporate sellers, the substantial shareholdings exemption (SSE) is often assumed to deliver a tax-free exit on a share sale. The SSE exempts from capital gains tax the profits on the sale of stakes of more than 10% under certain conditions. In the case of AI companies, that assumption requires careful validation. SSE requires the target to be a “trading” (or “operating” in the U.S.) company or the holding company of a trading/operating group.
AI businesses deploying models as part of a their customer-facing business should generally meet that test. However, where revenue is primarily derived from licensing models, granting access to datasets or centralized IP exploitation structures, there is a risk that nontrading activities become substantial — particularly at the group level, where offshore IP entities or licensing hubs are common.
This is not a black-or-white test, and many IP-rich businesses will still qualify. But the analysis is highly fact-sensitive. Where SSE does not apply, the tax cost (tax on the shareholding gains) is immediate and material. Sellers should be preparing this analysis well in advance of exit.
Incentives: Valuable, but Conditional
Two of the U.K.’s most valuable innovation incentives — R&D tax credits and the Patent Box (which promises an effective 10% tax rate on qualifying patent income) — often feature in a seller’s assessment of its own value. Both deserve a robust analysis in an AI context.
R&D claims built on model training and development can be legitimate, but they must meet the “advance in science or technology” test, and the company’s documentation must stand up to scrutiny. HM Revenue & Customs has signaled increased scrutiny of all R&D claims, including AI-sector claims. A buyer that inherits a history of generous R&D claims without robust supporting records is inheriting a risk, not necessarily a benefit.
The Patent Box only applies where there is an actual granted patent driving the revenue. Many AI businesses have no patents. Their edge is in data, architecture and execution, none of which qualify. The incentive is powerful where it applies, but it applies less often in AI than founders and investors tend to assume.
The Border Problem: When the AI Is the Agent
Here is the question that keeps tax advisers on AI M&A deals up at night: Suppose an AI agent, hosted on an overseas server, autonomously negotiates terms with U.K. customers and closes transactions with minimal (or no) human intervention. Has it just created a U.K. taxable presence for its owner?
These questions remain untested in the AI context, and any analysis will turn on broader questions of control and deployment of the AI agent. (To what extent is it really acting autonomously?)
However, the direction of travel is clear: As systems become more autonomous, the risk of creating an unexpected tax presence (i.e., bringing a non-U.K. company within the U.K. corporation tax net) increases, particularly where human oversight is limited.
Due diligence should now include a review of what the technology does, not just where the human team sits.
What This Means for Dealmakers
AI M&A is not simply a variant of software transactions. Buyers that do not interrogate the IP, test the operating status of the company and map the operational footprint will find the gap between headline valuation and after-tax return wider than expected.
This memorandum is provided by Skadden, Arps, Slate, Meagher & Flom LLP and its affiliates for educational and informational purposes only and is not intended and should not be construed as legal advice. This memorandum is considered advertising under applicable state laws.